| 1 | //go:build linux |
| 2 | |
| 3 | package console |
| 4 | |
| 5 | import ( |
| 6 | "context" |
| 7 | "encoding/json" |
| 8 | "errors" |
| 9 | "fmt" |
| 10 | "os" |
| 11 | "os/exec" |
| 12 | "strconv" |
| 13 | "strings" |
| 14 | "testing" |
| 15 | "time" |
| 16 | |
| 17 | "github.com/carlosframework/platform/internal/hostactors" |
| 18 | "github.com/carlosframework/platform/internal/hostfloor" |
| 19 | "github.com/carlosframework/platform/internal/release" |
| 20 | "github.com/carlosframework/platform/internal/status" |
| 21 | ) |
| 22 | |
| 23 | func TestPlan9Fix1ConsoleRestartReplacesDeadProcess(t *testing.T) { |
| 24 | bootRaw, err := os.ReadFile("/proc/sys/kernel/random/boot_id") |
| 25 | if err != nil { |
| 26 | t.Fatal(err) |
| 27 | } |
| 28 | boot := strings.TrimSpace(string(bootRaw)) |
| 29 | child := exec.Command("sleep", "30") |
| 30 | if err := child.Start(); err != nil { |
| 31 | t.Fatal(err) |
| 32 | } |
| 33 | t.Cleanup(func() { _ = child.Process.Kill() }) |
| 34 | stat, err := os.ReadFile(fmt.Sprintf("/proc/%d/stat", child.Process.Pid)) |
| 35 | if err != nil { |
| 36 | t.Fatal(err) |
| 37 | } |
| 38 | fields := strings.Fields(string(stat[strings.LastIndexByte(string(stat), ')')+1:])) |
| 39 | namespace, err := hostReportPIDNamespace() |
| 40 | if err != nil { |
| 41 | t.Fatal(err) |
| 42 | } |
| 43 | oldID := fmt.Sprintf("%s:%s:%d:%s", boot, namespace, child.Process.Pid, fields[19]) |
| 44 | nextID := hostReportProcessID(boot) |
| 45 | old := plan9Report(oldID, time.Now().Add(-time.Minute)) |
| 46 | next := plan9Report(nextID, time.Now()) |
| 47 | if err := hostReportCollision(old, next, time.Now()); err == nil { |
| 48 | t.Fatal("live predecessor allowed replacement") |
| 49 | } |
| 50 | _ = child.Process.Kill() |
| 51 | _ = child.Wait() |
| 52 | if err := hostReportCollision(old, next, time.Now()); err != nil { |
| 53 | t.Fatalf("restart blocked by dead predecessor: %v", err) |
| 54 | } |
| 55 | for _, id := range []string{"other:" + strconv.Itoa(child.Process.Pid) + ":1", boot + ":" + strconv.Itoa(child.Process.Pid) + ":bad"} { |
| 56 | if err := hostReportCollision(plan9Report(id, time.Now()), next, time.Now()); err == nil { |
| 57 | t.Fatalf("unproven predecessor accepted: %s", id) |
| 58 | } |
| 59 | } |
| 60 | } |
| 61 | |
| 62 | type plan9ReportReleaseStore struct { |
| 63 | release.Store |
| 64 | rosterReads int |
| 65 | failWrite bool |
| 66 | } |
| 67 | |
| 68 | func (s *plan9ReportReleaseStore) Get(ctx context.Context, key string) ([]byte, string, error) { |
| 69 | if key == hostactors.Key { |
| 70 | s.rosterReads++ |
| 71 | } |
| 72 | return s.Store.Get(ctx, key) |
| 73 | } |
| 74 | func (s *plan9ReportReleaseStore) Put(ctx context.Context, key string, b []byte, etag string) (string, error) { |
| 75 | if s.failWrite && strings.HasPrefix(key, "status/consoles/") { |
| 76 | return "", errors.New("report unavailable") |
| 77 | } |
| 78 | return s.Store.Put(ctx, key, b, etag) |
| 79 | } |
| 80 | func TestPlan9Fix1ReportFailureStillReleasesActivation(t *testing.T) { |
| 81 | for _, failure := range []string{"write", "collision"} { |
| 82 | t.Run(failure, func(t *testing.T) { |
| 83 | s, d, target := activationFixture(t) |
| 84 | p, err := s.promoteHostUpdate(t.Context(), target, "router", "stable", release.SystemPromoteOptions{}) |
| 85 | if err != nil || p.ActivationID == "" { |
| 86 | t.Fatal(p, err) |
| 87 | } |
| 88 | observed, err := d.Read(t.Context()) |
| 89 | if err != nil { |
| 90 | t.Fatal(err) |
| 91 | } |
| 92 | observed.Roster.Consoles["console-2"] = hostactors.Console{Enrollment: strings.Repeat("c", 32), InstanceID: "console-2", Namespaces: []string{"platform"}, DeploymentBucket: d.Bucket} |
| 93 | raw, err := release.SignRootEnvelope(release.RootSigner(s.cfg.Key), observed.Roster) |
| 94 | if err != nil { |
| 95 | t.Fatal(err) |
| 96 | } |
| 97 | if _, err = d.Store.Put(t.Context(), hostactors.Key, raw, observed.ETag); err != nil { |
| 98 | t.Fatal(err) |
| 99 | } |
| 100 | s.cfg.HostConsoleID, s.cfg.ReleasePub = "console-2", d.PublicKey |
| 101 | previous := hostRunningImage |
| 102 | hostRunningImage = func() (hostfloor.Running, error) { return hostfloor.Running{BootID: "boot"}, nil } |
| 103 | t.Cleanup(func() { hostRunningImage = previous }) |
| 104 | if failure == "collision" { |
| 105 | b, _ := json.Marshal(hostactors.Report{ProcessID: "foreign:22", GeneratedAt: time.Now()}) |
| 106 | if _, err := d.Store.Put(t.Context(), "status/consoles/console-2.json", b, ""); err != nil { |
| 107 | t.Fatal(err) |
| 108 | } |
| 109 | } |
| 110 | boxRaw, boxETag, err := d.Store.Get(t.Context(), "status/boxes/box.json") |
| 111 | if err != nil { |
| 112 | t.Fatal(err) |
| 113 | } |
| 114 | var box status.Box |
| 115 | if err := json.Unmarshal(boxRaw, &box); err != nil { |
| 116 | t.Fatal(err) |
| 117 | } |
| 118 | box.Metrics.Release, box.Metrics.ReleaseTag = target.Version, target.ReleaseTag |
| 119 | box.Metrics.BinarySHA256 = target.Artifacts["linux-amd64"].SHA256 |
| 120 | box.Metrics.ActivationID, box.Metrics.ActivationComplete = p.ActivationID, true |
| 121 | boxRaw, err = json.Marshal(box) |
| 122 | if err != nil { |
| 123 | t.Fatal(err) |
| 124 | } |
| 125 | if _, err := d.Store.Put(t.Context(), "status/boxes/box.json", boxRaw, boxETag); err != nil { |
| 126 | t.Fatal(err) |
| 127 | } |
| 128 | spy := &plan9ReportReleaseStore{Store: d.Store, failWrite: failure == "write"} |
| 129 | s.cfg.Releases = spy |
| 130 | ctx, cancel := context.WithTimeout(t.Context(), 100*time.Millisecond) |
| 131 | defer cancel() |
| 132 | s.StartHostReports(ctx) |
| 133 | after, err := d.Read(t.Context()) |
| 134 | if err != nil { |
| 135 | t.Fatal(err) |
| 136 | } |
| 137 | if _, held := after.Roster.Activations[p.ActivationID]; held { |
| 138 | t.Fatal("completed activation remains reserved after report failure") |
| 139 | } |
| 140 | if spy.rosterReads < 3 { |
| 141 | t.Fatalf("ReleaseActivation was skipped after %s failure: roster reads=%d", failure, spy.rosterReads) |
| 142 | } |
| 143 | }) |
| 144 | } |
| 145 | } |
| 146 | |